Abstract
With written Swiss German becoming more popular in everyday use, it has become a target for text processing. The absence of a standard orthography and the variety of dialects, however, lead to a vast variation in different spellings which makes this task difficult. We built a system based on weighted transducers that recognizes over 90% of the tokens in certain texts. Weights ensure preferring the best analysis for most words while at the same time allowing for very broad range of spelling variations. Our morphological tagset that we defined for this purpose and lemmas in Standard German open the possibility for further processing. Besides our morphological analyzer and lemmatizer, a morphologically annotated corpus offers new resources for Swiss German and helps spreading our tagset.